Oh, yes, I guess Radul's information propagation is related. Aside from the state model, I think Tenerezza and that system could be seen as competitor dialects with some minor technical and philosophical quirks keeping them distinct from each other, like Ruby and Python.
Radul talks about dealing with partially known values, and I don't exactly have partial values in Tenerezza; I just have sets, and a partial set is a set.
Radul talks about the notion of a partially known value that improves with time, and that notion is inherently stateful. But Radul's state nodes are second-class, and for some reason they're intermingled with stateless propagation nodes. I first saw that paper after I had already been hacking along to David Barbour's vision of programming where there are first-class interfaces to state but the state always resides outside the running program. I'm not sure I see the appeal of second-class state nodes, and on the flip side, I'm not sure I see the point of having stateless propagation nodes if the rest of the network can be stateful.
For Tenerezza, I haven't tackled state yet, but I have an approach in mind. I think it's very different from Radul's approach, but maybe not different enough that the Ruby/Python comparison breaks down.
In Tenerezza, state can be accessed via first-class channels, which is naturally justified by the metaphor that you can ask a chat buddy to remember something for you. However, every Tenerezza call has access to a "conscience" channel that appears out of nowhere so that the program can explicitly interact with the language runtime itself. This channel is naturally justified by the fact that the language runtime has the power to disrupt and interact with the program whether the programmer wants it to or not. I call it the conscience because I expect it to be a good channel to consult for advice if there's an error, and it doubles as a good channel to supplicate for call-specific state resources.
I have a very vague plan to support some kind of generic operation for resource reallocation, which would be useful for managing the allocation of existing state resources, randomness resources, robots, and so on. The premise would be that if the two sides of a communication channel pitch in resources and simultaneously agree to the same protocol, then that agreement itself becomes a third entity that the two can communicate with. We understand our world through a process of social reinterpretation, so I'm excited to think that this social reinterpretation operator would singlehandedly be a sufficient protocol to interact with whatever we encounter in the world.
---
"That increment example looks like just the thing. Looking forward to it!"
Okay, I'll try to get that written up at some point. XD;
